Views
3 years ago

Evaluation of IREC's Statistical Analysis and Claims

Evaluation of IREC's Statistical Analysis and Claims

evidence from the

evidence from the consideration of the commissioners and the Kenyan people. Third, IREC’s research design – specifically its decision to examine 19 problematic constituencies – was not capable of determining whether or not rigging occurred at KICC. Fourth, given its resources and time-frame, IREC could have conducted research capable of answering the questions at hand; they either remained ignorant of the proper methods for doing so, or deliberately chose not to pursue an effective analytical strategy. We hope that this analysis will provide Kenyans with enough information to decide for themselves whether or not to accept IREC’s final report as accurate, or whether they have been short-changed by the process. 2

Sampling problems Problem 1 IREC adamantly and repeatedly stated that statistical analysis should not be carried out on the electoral data: “[T]he IREC analysis of tallying etc. in nineteen sample constituencies demonstrates convincingly that there are so many more or less erroneous constituency results [...] that one cannot rely on any figures from the ECK.” 2 Response 1 IREC’s claim that their findings from a sample of 19 constituencies implies that the rest of the constituency results are equally flawed. This claim cannot be supported with the methods that IREC used, and demonstrates a basic misunderstanding of simple statistical concepts. The reasoning underpinning IREC’s decision not to pursue statistical analysis demonstrates a misunderstanding of basic statistical concepts like random sampling, as well as an ignorance more sophisticated statistical election forensics. Their tenuous logic on why not to do electoral statistics proceeds in two steps. We examine each step in turn, assessing the validity of each step. First, IREC states that “[a]lmost all parliamentary and presidential election results for the constituencies sampled are erroneous, which means that very few of the officially published figures are actually accurate” (p. 127). Put another way, IREC is saying that because the ECK’s results in the 19 “sampled” constituencies contained many errors, most of the other 191 constituencies contain similar errors. A student with even the most basic statistical training will recognise this conclusion as false. The mistake that IREC made lies in the way it chose its sample. In short, because IREC’s sample focuses on disputed constituencies rather than a random sample of all constituencies, their findings in that sample cannot generalise to all 210 constituencies. Again, we quote the report in full to capture the magnitude of their error: The magnitude of problem that brought about the dispute in results from the 2007 elections is enormous. The entire result and a significant part of the election was in serious dispute. Therefore, the choice of what to analyze needed to be very strategic and robust, but at the same time careful enough in order not to leave out key areas that had serious disputes, or those that could be instrumental in helping the IREC achieve its mandate. In order to identify the constituencies for analysis we were guided by a number of criteria, namely: a. Looking at specific disputed constituencies listed by various parties/bodies. For instance, submissions sent to IREC by Orange Democratic Movement (ODM), Party of National Unity (PNU) and other bodies, listed a number of problematic constituencies: [sic] b. Selecting areas using a structured selection criteria based on various electoral anomalies such alleged high-turn out and so on [sic] 2 IREC report, p. 136. 3

Statistical concepts in survival analysis - IARC
Evaluation of precipitation statistics in the CLM and sensitivity to ...
Statistical Design and Analysis - BEBAC • Consultancy Services for ...
Statistical design, analysis and interpretation of gene expression ...
Statistical Analysis of proteomic data
Overview of Evaluations of Large-Scale Statistical ... - Paris21
Statistical Data Analysis - Supreme Court of Pakistan
M t i d St ti ti l Metrics and Statistical Analysis in Test y and Evaluation
official Statistics and Monitoring and Evaluation Systems in ... - Paris21
Statistical Analysis of Cell-Mediated Statistical ... - GLOBE Network
SASIE System Analysis, Simulation and Evaluation
Statistical analysis of the water quality of the Fitzroy ... - OzCoasts
australian small business key statistics and analysis - Department of ...
Statistical Analysis Fundamentals Statistical Analysis Fundamentals ...
Product Sensory Evaluation for Claim Support - Sensory Spectrum
Statistical analysis of differences in liking of Napoli ... - Agrostat 2010
WACOSS analysis of the ORIMA Research Evaluation of the Child ...
I. Cost-Benefit Analysis and Impact Evaluation for ... - Euromedina
spatial analysis + policy evaluation - School of Environment and ...
Certificate in Job Analysis & Descriptions, Job Evaluation - jobmaster
Project Analysis and Evaluation - Bureau of Land Management
A Statistical Analysis of Ringfort Distribution and Morphology on the ...
Statistical analysis of wave activity in ionosphere from 2003-2012 ...
Evaluability Assessment - Justice Research and Statistics Association
Strategies for Evaluating - Justice Research and Statistics Association
Pharmacokinetic and Statistical Analysis of BE Data - BEBAC ...
Social Network Analysis and Evaluation - Environmental Evaluators ...